问题现状

在Java实现批量下载多文件(夹)压缩包(zip)篇幅中通过在服务器上创建临时文件，借助hutool的ZipUtil将文件(夹)压缩写入至response的OutputStream，实现了多文件(夹)的压缩包下载。其大致流程图可大致描述为：

经过分析和验证上述方式实现的批量下载存在着下列问题

• 1.文件非常大的情形下，步骤1.2. 4将文件先下载到服务器带来了额外的耗时操作，对于用户来说下载文件只需要将文件从文件系统直接写入响应即可。
• 2.由于请求类型为POST，所以浏览器不能自动下载文件，步骤5即使将流已写入响应，但是浏览器并不能打开下载页面，需要前端接收到所有Blob才能打开下载，用户体验极差，易给用户造成批量下载没反应的错觉。

是否存在一种方案，可以将批量下载接口转为GET请求，且可以将文件(夹)直接写入到response的OutputStream？

解决思路

1.首先由于批量下载接口batchDownloadFile的参数类型为List<DownloadFileParam>为复杂参数，故无法直接将POST请求修改为GET；这时候该怎么办呢？

架构思维中，比较常用的一种思路便是分层架构！我门可以将批量下载接口拆为两个接口

通过POST方式保存下载参数List<DownloadFileParam>到Redis，并返回Redis中该下载参数对应唯一标示key的接口getBatchDownloadKey如下

@PostMapping(value = "/getBatchDownloadKey")
public String getBatchDownloadKey(@RequestBody List<DownloadFileParam> params)...

根据返回下载参数唯一标示Key进行批量下载的GET接口batchDownloadFile接口，定义如下

@GetMapping(value = "/batchDownloadFile", produces = "application/octet-stream)
public void batchDownloadFile(@RequestParam("downLoadKey") String downLoadKey)

2.Java提供了类ZipArchiveOutputStream允许我们可以直接将带有目录结构的文件压缩到OutputStream，其使用的伪代码如下

ZipArchiveOutputStream zous = new ZipArchiveOutputStream(response.getOutputStream());
//file为带有目录结构的文件比如:/文件夹/子文件夹/文件.txt
ArchiveEntry entry = new ZipArchiveEntry(file);
InputStream inputStream = file.getInputStream();
zous.putArchiveEntry(entry);
try {
  int len;
  byte[] bytes = new byte[1024];
  //inputStream为文件流
  while ((len = inputStream.read(bytes)) != -1) {
    zous.write(bytes, 0, len);
  }  
  zous.closeArchiveEntry();
  zous.flush();
  } catch (Exception e) {
    e.printStackTrace();
  } finally {
    IoUtil.close(inputStream);
  }

这样我们就可以避免将文件下载到服务器带来的性能消耗。

3.整个过程的流程图如下

代码实现

保存下载参数请求getBatchDownloadKey

@PostMapping(value = "/getBatchDownloadKey")
public String getBatchDownloadKey(@RequestBody List<DownloadFileParam> params) throws Exception {
    try {
        String key = IdGenerator.newShortId();
        redisTemplate.opsForValue().set(key, JSONObject.toJSONString(params), 60, TimeUnit.SECONDS);
        return key;
    } catch (Exception e) {
        logger.error("getBatchDownloadKey error params={}", params, e);
        throw e;
    }
}

根据Key下载文件的接口定义batchDownloadFile

@GetMapping(value = "/pass/batchDownloadFile", produces = "application/octet-stream;charset=UTF-8")
public void batchDownloadFile(@RequestParam("downLoadKey") String downLoadKey,@RequestParam("token") String token) throws Exception {
    try {
        fileService.batchDownloadFile(downLoadKey, getRequest(), getResponse(),token);
    } catch (Exception e) {
        logger.error("batchDownloadFile error params={}", downLoadKey, e);
        throw e;
    }
}

fileService.batchDownloadFile

@Override
    public void batchDownloadFile(String key, HttpServletRequest request, HttpServletResponse response,String token) throws Exception {
        if (redisUtil.get(token) != null) {
            UserSession userSession = JSONObject.parseObject(redisUtil.get(token).toString(), UserSession.class);
            //如果存在session或者token是存在于project_token配置的值，通过认证
            if (userSession != null) {
                Object result = redisTemplate.opsForValue().get(key);
                if (result == null) {
                    throw new ParamInvalidException("无效的批量下载参数key");
                }
                List<DownloadFileParam> params = JSONArray.parseArray(result.toString(), DownloadFileParam.class);
                //创建虚拟文件夹
                String mockFileName = IdGenerator.newShortId();
                String tmpDir = "";
                FileUtil.mkdir(tmpDir);
                ZipArchiveOutputStream zous = null;
                try {
                    //设置响应
                    response.reset();
                    response.setContentType("application/octet-stream");
                    response.setHeader("Accept-Ranges", "bytes");
                String fileName = URLEncoder.encode(DateFormatUtil.formatDate(DateFormatUtil.yyyyMMdd, <span class="hljs-keyword">new</span> Date()) + <span class="hljs-string">".zip"</span>, <span class="hljs-string">"UTF-8"</span>).replaceAll(<span class="hljs-string">"\\+"</span>, <span class="hljs-string">"%20"</span>);
                response.setHeader(<span class="hljs-string">"Content-disposition"</span>, <span class="hljs-string">"attachment;filename*=utf-8''"</span> + fileName);
                response.setHeader(<span class="hljs-string">"Access-Control-Expose-Headers"</span>, <span class="hljs-string">"Content-Disposition"</span>);
                <span class="hljs-comment">//参数组装</span>
                zous = <span class="hljs-keyword">new</span> ZipArchiveOutputStream(response.getOutputStream());
                zous.setUseZip64(Zip64Mode.AsNeeded);

                DownloadFileParam downloadFileParam = <span class="hljs-keyword">new</span> DownloadFileParam();
                downloadFileParam.setFileName(mockFileName);
                downloadFileParam.setIsFolder(<span class="hljs-number">1</span>);
                downloadFileParam.setChilds(params);

                <span class="hljs-comment">//递归文件流添加zip</span>
                downloadFileToServer(tmpDir, downloadFileParam, zous);
                zous.closeArchiveEntry();
            } <span class="hljs-keyword">finally</span> {
                zous.close();
            }
        } <span class="hljs-keyword">else</span> {
            <span class="hljs-keyword">throw</span> <span class="hljs-keyword">new</span> ResultException(<span class="hljs-string">"服务内部错误"</span>);
        }
    } <span class="hljs-keyword">else</span> {
        <span class="hljs-keyword">throw</span> <span class="hljs-keyword">new</span> ResultException(<span class="hljs-string">"用户已下线,请重新登录"</span>);
    }
}

downloadFileToServer

private void downloadFileToServer(String tmpDir, DownloadFileParam downloadFileParam, ZipArchiveOutputStream zous) throws Exception {
    List<DownloadFileParam> childs = downloadFileParam.getChilds();
    if (EmptyUtils.isNotEmpty(childs)) {
        final String finalPath = tmpDir;
        childs.stream().forEach(dwp -> dwp.setFile(EmptyUtils.isNotEmpty(finalPath) ? finalPath + File.separator + dwp.getFileName() : dwp.getFileName()));
        for (int i = 0; i < childs.size(); i++) {
            DownloadFileParam param = childs.get(i);
            if (param.getIsFolder() == 0) {
                FileInfo fileInfo = fileInfoDao.findById(param.getFileId()).orElseThrow(() -> new DataNotFoundException("文件不存在或已被删除!"));
                List<GridFsResource> gridFSFileList = fileChunkDao.findAll(fileInfo.getFileMd5());
                ArchiveEntry entry = new ZipArchiveEntry(param.getFile());
                zous.putArchiveEntry(entry);
                if (gridFSFileList != null && gridFSFileList.size() > 0) {
                    try {
                        for (GridFsResource gridFSFile : gridFSFileList) {
                            InputStream inputStream = gridFSFile.getInputStream();
                            try {
                                int len;
                                byte[] bytes = new byte[1024];
                                while ((len = inputStream.read(bytes)) != -1) {
                                    zous.write(bytes, 0, len);
                                }
                            } finally {
                                IoUtil.close(inputStream);
                            }
                        }
                        zous.closeArchiveEntry();
                        zous.flush();
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            }
            //递归下载文件到压缩流
            downloadFileToServer(tmpDir, param, zous);
        }
    }
}

方案总结

一般情况下下载接口最好用GET方式，浏览器会自动开始下载，除此之外，接口参数与下载接口参数间通过添加中间层解藕帮我们解决了POST下载转化为GET下载方式的问题，分层的架构思想是软件架构最常用的一种方式，再解决工作实际问题的过程中，我们要善于变通采用该方式。